I join the opinion of the court because I agree that it is unworkable to strike down the authorization rule articulated in Commonwealth v. Santana, 420 Mass. 205, 649 N.E.2d 717 (1995). However, I write separately because, although-as the court points out-the driver here was not stopped for "driving while black," it is important to highlight how pretextual stops disproportionately affect people of color, and to explore what can be done to mitigate the harm caused by this practice.
Years of data bear out what many have long known from experience: police stop drivers of color disproportionately more often than Caucasian drivers for insignificant violations (or provide **877no reason at all). In 2017, the Stanford Open Policing Project found that police stopped African-American drivers more than Caucasian drivers, controlling for population makeup, both nationally and in Massachusetts.1 Stanford Open Policing, Stop Rates, 2017, https://openpolicing.stanford.edu/findings/ [https://perma.cc/F6HT-87WE]. See United States Department of Justice, Office of Justice Programs, Bureau of Justice Statistics, Special Report, Police Behavior During Traffic and Street Stops, 2011, at 3 (rev. October 27, 2016), https://www.bjs.gov/content/pub/pdf/pbtss11.pdf [https://perma.cc/2ML3-UWY9].
In effectuating traffic stops, most officers act in good faith. Even where they do, to a Caucasian driver a traffic stop may be annoying or embarrassing, but for a driver of color, such a stop can be humiliating and painful.2 Commonwealth v. Feyenord, 445 Mass. 72, 88, 833 N.E.2d 590 (2005), cert. denied, 546 U.S. 1187, 126 S.Ct. 1369, 164 L.Ed.2d 77 (2006) (Greaney, J., concurring). Further, recent tragic events have shown that the fear people of color have of being stopped by police is justified: African-Americans have been killed during routine traffic stops.3
*782**878It goes without saying that this is not a new phenomenon. Almost twenty years ago, then-Associate Justice Ireland noted statistics from multiple jurisdictions showing that African-American and sometimes Hispanic drivers were stopped more often than Caucasian drivers, even though Caucasian drivers were the majority group. Commonwealth v. Gonsalves, 429 Mass. 658, 670, 711 N.E.2d 108 (1999) (Ireland, J., concurring).
The reasons for pretextual stops of people of color stem from explicit bias (i.e., racial profiling), unconscious bias,4 or a combination of both. See Carbado, From Stopping Black People to Killing Black People: The Fourth Amendment Pathways to Police Violence, 105 Cal. L. Rev. 125, 129-130 (2017) ; Harris, The Stories, the Statistics, and the Law: Why "Driving While Black" Matters, 84 Minn. L. Rev. 265, 291-292 (1999) ; Ramirez, Hoopes, & Quinlan, Defining Racial Profiling in a Post-September 11 World, 40 Am. Crim. L. Rev. 1195, 1197-1198 (2003). See also Greenwald & Krieger, Implicit Bias: Scientific Foundations, 94 Cal. L. Rev. 945, 951 (2006) ; Lawrence, The Id, the Ego, and Equal Protection: Reckoning with Unconscious Racism, 39 Stan. L. Rev. 317, 343 (1987). Regardless of the cause, it is a persistent, pervasive problem that must be addressed.
The solution, however, is not clear cut. For the reasons outlined by the court, the answer is not to overrule the authorization rule articulated in Santana, 420 Mass. at 208-209, 649 N.E.2d 717. As the court has explained, inquiring into subjective police intent for traffic stops would lead to several practical difficulties, not least among them **879the question of how precisely to determine intent. Ante at 865-866, 90 N.E.3d at 772-774.
In Commonwealth v. Lora, 451 Mass. 425, 886 N.E.2d 688 (2008), the court reiterated that although "law enforcement officers enjoy considerable discretion in exercising some selectivity for purposes consistent with the public interest," that "selectivity" cannot be based on "an unjustifiable standard such as race, religion or other arbitrary classification"5 (quotations and citations omitted). Id. at 436-437, 886 N.E.2d 688. The court concluded that to rebut the presumption that a stop was not undertaken as a result of an arbitrary *783classification, a defendant must present "credible evidence establishing a reasonable inference of impermissible discrimination." Id. at 443, 886 N.E.2d 688. The court further held that
"[a]t a minimum, that evidence must establish that the racial composition of motorists stopped for motor vehicle violations varied significantly from the racial composition of the population of motorists making use of the relevant roadways, and who therefore could have encountered the officer or officers whose actions have been called into question."
Id. at 442, 886 N.E.2d 688.
Thus, the court attempted to provide a means of combatting pretextual stops based on race with statistics. We noted that a similar approach had been somewhat successful in New Jersey. Id. at 440-441, 886 N.E.2d 688, citing State v. Soto, 324 N.J. Super. 66, 734 A.2d 350 (1996). As it happened, traffic stop statistics also were being collected in the Commonwealth. Before Lora was decided, the Legislature had passed An Act providing for the collection of data relative to traffic stops (act), St. 2000, c. 228. Pursuant to the act, Northeastern University analyzed a year's worth of data collected on racial and gender profiling, and issued a report in 2004. Lora, 451 Mass. at 448, 886 N.E.2d 688. Despite the Legislature's focus on data collection in this act, the court acknowledged that the defendant's evidentiary burden was "daunting." Id. at 445, 886 N.E.2d 688.
In a concurring opinion, then-Justice Ireland pointed out some of the difficulties involved in collecting the necessary data, even with the act in place. Id. at 449, 886 N.E.2d 688 (Ireland, J., concurring). For example, although the act required law enforcement agencies that **880had racially profiled to continue to gather statistics, it did not contain provisions requiring those agencies to report the data to anyone or to analyze the data, severely undercutting any use that data might have had. Id. (Ireland, J., concurring). Moreover, almost one-half of the targeted agencies failed to follow the reporting guidelines of the act, for example by failing to track certain factors or failing to report at all. Id. (Ireland, J., concurring).
Justice Ireland's concerns were prescient: the act required governmental data collection for only a limited amount of time, and the Legislature has not renewed the necessary funding. See St. 2000, c. 228, § 8 (assigning financial responsibility to State agencies); id. at § 10 (requiring data to be transmitted for analysis after one year). Statistics on traffic stops, thus, are now even more difficult to come by. We are not aware of any traffic stop cases in which a defendant has been able to gather and use statistics to prove that the stop violated equal protection principles; it appears that Lora has not provided the opportunity for defendants that we had hoped it would.
Concerns about bias in pretextual traffic stops are well founded, as are concerns about the practical ability of defendants to show racial bias by way of statistics as suggested by Lora. Because this is not a "driving while black" equal protection case, the issue is not squarely before us. However, it is worth noting that it has been seventeen years since the Legislature required State agencies to collect data on racial profiling. We are not aware of the data ever being used to mount a challenge under Lora, and it is now woefully outdated. The time has come for the Legislature to address the problem once more. Publicly available data would not only assist litigants, but would also inform the public about this ongoing problem.
*784In the meantime, our recent holding in Commonwealth v. Cordero, 477 Mass. 237, 74 N.E.3d 1282 (2017), has added to our jurisprudence. There we held that a traffic stop may go no further than investigating the alleged traffic violation unless that investigation leads to information to support reasonable suspicion of a crime. Id. at 247, 74 N.E.3d 1282. See Commonwealth v. Amado, 474 Mass. 147, 151, 48 N.E.3d 414 (2016) ; Gonsalves, 429 Mass. at 663, 711 N.E.2d 108 ; Commonwealth v. Torres, 424 Mass. 153, 158-159, 674 N.E.2d 638 (1997). These cases are by no means a cure for racial profiling in traffic stops, but they may provide a means to lessen their impact on drivers and diminish the incentive to conduct pretextual stops.
I note that although most of the data focuses on people of color, other marginalized communities, i.e., groups of people who have historically experienced some form of oppression or exclusion, are also the target of heightened police attention. Transgendered people, for example, have reported facing disproportionate harm by encounters with law enforcement. Activists Say Police Abuse of Transgender People Persists Despite Reforms, New York Times, Sept. 6, 2015.
In Commonwealth v. Warren, 475 Mass. 530, 540, 58 N.E.3d 333 (2016), when we discussed the related problem of racial profiling in Terry-type stops, we noted "the recurring indignity of being racially profiled."
The following are a few recent examples that have gained national attention. A police officer in Minnesota stopped Philando Castile for a broken taillight. During the encounter, the officer shot him four times, killing him in front of his fiancée and four year old daughter. Woman Streams Aftermath of Fatal Officer-Involved Shooting, Cable News Network, July 8, 2016, http://www.cnn.com/2016/07/07/us/falcon-heights-shootingminnesota/index.html [https://perma.cc/4P5A-YY28]. In Ohio, the police stopped Samuel DuBose for failing to display a front license plate, and fatally shot him during the stop. The Shooting of Samuel DuBose, New York Times, July 29, 2015. The South Carolina police stopped Walter Scott for a broken taillight, and shot him to death as he fled. Carbado, From Stopping Black People to Killing Black People: the Fourth Amendment Pathways to Police Violence, 105 Cal. L. Rev. 125, 149 (2017). In Texas, a police officer stopped Sandra Bland for failing to signal a lane change. Id. at 150. She was found dead in jail three days later. Id.
Massachusetts is not immune from traffic stop violence. Wakeelah Cocroft, an African-American woman, was a passenger in a vehicle that the police stopped for speeding in Worcester. Cocroft v. Smith, 95 F.Supp.3d 119, 123 (D. Mass. 2015). During the stop, an officer "forcefully threw Cocroft to the ground and scraped her face against the cement." Id. In a subsequent civil suit, a jury found that the officer had unlawfully seized Cocroft. Id. at 122.
It is also important to note that these examples are not meant to diminish the fact that police officers are at risk during traffic stops as well. Auburn police officer Ronald Tarentino, for example, was shot to death during a traffic stop. Obituary for Fallen Police Officer Ronald Tarentino, Jr., Boston Herald, May 24, 2016, http://www.bostonherald.com/news/local_coverage/herald_bulldog/2016/05/obituary_for_fallen_police_officer_ronald_tarentino_jr [https://perma.cc/8GNT-KQRU].
Unconscious or implicit bias is a discriminatory belief or association likely unknown to its holder. Multiple studies confirm the existence of implicit bias, and that implicit bias predicts real-world behavior. See Kang & Banaji, Fair Measures: A Behavioral Realist Revision of "Affirmative Action," 94 Cal. L. Rev. 1063, 1071-1073 (2006). That is, even people who do not believe themselves to harbor implicit bias may in fact act in ways that disfavor people of color.
As the court points out, the defendant did not bring a claim under the equal protection provisions of the Massachusetts Constitution, another fatal blow to mounting a challenge to pretextual stops. Ante at 870-871, 90 N.E.3d at 776-777.